armel is already compiled and supported upstream by the Debian project. I agree, Raspbian is fantastic and I'm amazed by what mpthompson, plugwash and others have achieved. So don't worry, there's no time spent compiling (at least, for those packages already available in Debian). For the reasons you state, it may indeed make a lot of sense to move to the raspbian hardfloat port in the future.
I think in the interest of producing the most optimised version of Debian for the Pi going forward a switch to Raspbian will only make sense for the foundation.
A number of tests have been conducted that clearly show the benefit of what has been achieved by mptompson, plugwash etc.
On limited hardware such as the Pi every possible optimisation counts. By switching to Raspbian you provide a stable target before the Pi hardware ships to the larger educational sector.
Developers can target the Floating operations this release provides and really improve their applications and thus improve the whole of the Raspberry Pi experience for the (hopefully) millions of users that will receive this little credit card sized miracle! It can only make sense to choose the best and most optimised software available as the base for the Pi.
I can understand why an Armel version might be useful, as explained by Plugwash (as a bug handling mechanism) but it creates incompatible versions of the software and division and further work. If the effort of the foundation is concentrated on Raspbian then amazing progress will inevitably be made and developers will know what to start writing against.